Forecasting GICs and Geoelectric Fields From Solar Wind Data Using LSTMs: Application in Austria

The forecasting of local GIC effects has largely relied on the forecasting of dB/dt as a proxy and, to date, little attention has been paid to directly forecasting the geoelectric field or GICs themselves. We approach this problem with machine learning tools, specifically recurrent neural networks or LSTMs by taking solar wind observations as input and training the models to predict two different kinds of output: first, the geoelectric field components Ex and Ey; and second, the GICs in specific substations in Austria. The training is carried out on the geoelectric field and GICs modeled from 26 years of one‐minute geomagnetic field measurements, and results are compared to GIC measurements from recent years. The GICs are generally predicted better by an LSTM trained on values from a specific substation, but only a fraction of the largest GICs are correctly predicted. This model has a correlation with measurements of around 0.6, and a root‐mean‐square error of 0.7 A. The probability of detecting mild activity in GICs is around 50%, and 15% for larger GICs. Plain Language Summary Using satellites, we measure the state of the solar wind a short distance away from the Earth (at the so‐called Lagrange‐1 or L1 point) to see what is coming toward us at any given moment. Changes in the solar wind such as an increase in wind speed or a strong magnetic field can potentially impact satellite operation in orbit and power grid infrastructure on the ground ‐ in extreme cases, solar storms can damage power grids and transformers by inducing electrical currents in the power lines. These are called geomagnetically induced currents (GICs). Here, we attempt to forecast the scales of GICs by applying machine learning methods, specifically Long‐Short‐Term‐Memory recurrent neural networks, to take the solar wind data measured at the L1 point and predict the currents that would be seen in power grids in Austria. This gives us a lead time of around 10–40 min in the forecast. We discuss whether it is best to attempt to predict the ground electric field that leads to the GICs or the GICs themselves, and discuss the difficulties in this kind of prediction and the shortfalls in the model. Key Points The aim is to directly forecast Geomagnetically induced currents (GICs) rather than $dB/dt$, which is often used as a proxy Results from LSTMs predicting either Ex and Ey or substation GICs from solar wind data are compared GIC forecasting seems to work best when the LSTM model is traine